Replication Terms and Concepts

Namespace refers to the directory structure which contains replicated data. Replicated data is always transferred separately from namespace data (although some small file data is transferred along with the namespace).

Namespace realization refers to the process in which the replicated directory structure (the namespace) appears on the replication target.

Because file data and namespace data is transferred separately, in some situations it might take longer for replicated data to complete transferring than for the namespace realization to complete. This is especially likely to happen if there is a backlog of file data waiting to be transferred at the time when namespace is either scheduled to run or is manually initiated.

The Blockpool is a data repository on the target. A blockpool is required on each machine used for replication or deduplication. If you use only replication, the blockpool file system can be small. If you configure deduplication as well as replication, the blockpool file system must be larger: at least large enough to hold the pool of deduplicated data segments.

When you configure StorNext for the first time, the Configuration Wizard enables you to specify the name of the file system you want to use for the blockpool.

Note: Once you specify the file system on which the blockpool resides, you cannot later choose a different blockpool file system. Use care when specifying the blockpool file system.

A Blackout is a period during which replication does not occur. You may schedule replication blackouts for periods in which you do not want replication data transfers to occur on a busy network. For example, an administrator may create a blackout during periods when WAN usage and traffic is the heaviest. In this situation replication might run after hours when WAN usage and traffic would be much lower.

A replication source policy is a replication/deduplication policy that has "Outbound Replication" turned On via the policy's Outbound Replication tab.

The policy also has a Source Directories tab. The directories specified on this tab will be replicated, and these directories are called replication source directories.

A replication target directory is the location to which replicated data is sent. The replication target may be a directory on a separate host machine, or it may be a directory on the source host machine. Regardless of where the target directory resides, it is very important that you use the replication target directory only for replicated data. Also, do not allows users to modify the files in the replication target directories.

When creating replication target directories, remember that the target directory must be at least as large as the sum of all replication source directories from which replicated data is sent. For example, if you have two source directories that are both 100GB, your replication target directory must be at least 200GB.

You can specify a replication schedule to define when the file system namespace realization should occur for an outbound replication schedule. For example, you might specify that you want namespace realization to occur at 6am and 6pm every day.

If you do not specify a replication schedule, you must manually run the replication policy whenever you want the realization to occur.

Replication Copies is the number of copies of replicated data saved on the target. StorNext currently supports 1 to 16 replication copies per target. The number of replication copies is entered or modified in replication policies.

Bandwidth Throttling refers to limiting the receive rate and transmit rate for replicated data (Replication Stage 1). This StorNext feature allows network administrators to specify (in bytes per second) a ceiling for incoming and outgoing replicated data. When bandwidth throttling is enabled, replicated data will not be transmitted or received at a rate higher than the specified maximum. Bandwidth throttling is a useful tool for controlling network traffic.

StorNext provides support for Multilink configurations, which means you can have multiple connections on one network interface card (NIC), or even multiple connections on multiple NICs. StorNext provides a tool that displays the NICs on your system, and allows you to specify the number of channels per NIC. On this same screen you can specify the NICs you want enabled for replication.

One advantage of using multiple NICs (or multiple channels on one NIC) is higher aggregate bandwidth because you can have multiple parallel paths. For this reason, multilink is valuable for load balancing.

When configuring multilink, be aware that the routing table of the host operating system determines the interface used for a connection to a specific endpoint. If there are multiple NICs in the same subnet as the destination IP endpoint, the host OS selects what it considers the best route and uses the interface associated with that route.

Note: An alternative to StorNext Multilink is to use the Linux bonding driver to enslave multiple Ethernet interfaces into a single bond interface. This is a Linux feature, not a StorNext feature, but is supported by the StorNext software.

Virtual IP or vIP is similar to an alias machine name. StorNext uses virtual IP addresses to communicate with machines rather than using the physical machine name. Virtual IPs are required in HA (high availability) environments, and are also used for multilink NICs.

Your network administrator can provide you with the virtual IP addresses and virtual netmasks you need for your system.

Note: If your replication source policy or target policy is an HA system, you must specify the vIP address in the field labeled “Address for Replication and Deduplication” on the Outbound Replication tab for the policy named "global" on each file system for which you will use replication. The default value for this field is "localhost".